XOR Neural Network in python using MLP Back-Propagation

XOR.py

import numpy as np

def sigmoid(x):
    return 1.0 / (1.0 + np.exp(-x))

def sigmoid_prime(x):
    return x * (1.0 - x)

epochs = 5000
input_size, hidden_size, output_size = 2, 3, 1
learning_rate = 0.1

# Truth table
X = np.array([[0, 0], [0, 1], [1, 0], [1, 1]])
Y = np.array([[0], [1], [1], [0]])

# Fill hidden and output layers with random values.
w_hidden = np.random.uniform(size=(input_size, hidden_size))
w_output = np.random.uniform(size=(hidden_size, output_size))

# Learning iteration
for epoch in range(epochs):
    # Forward propagation
    actual_hidden = sigmoid(np.dot(X, w_hidden))
    output = np.dot(actual_hidden, w_output)

    # Calculate error (expected output - calculated output)
    error = Y - output

    # Backward Propagation
    dZ = error * learning_rate
    w_output += actual_hidden.T.dot(dZ)

    dH = dZ.dot(w_output.T) * sigmoid_prime(actual_hidden)
    w_hidden += X.T.dot(dH)

    
actual_hidden = sigmoid(np.dot([0, 0], w_hidden))
actual_output = np.dot(actual_hidden, w_output)
print('[0, 0]', actual_output)

actual_hidden = sigmoid(np.dot([0, 1], w_hidden))
actual_output = np.dot(actual_hidden, w_output)
print('[0, 1]', actual_output)

actual_hidden = sigmoid(np.dot([1, 0], w_hidden))
actual_output = np.dot(actual_hidden, w_output)
print('[1, 0]', actual_output)

actual_hidden = sigmoid(np.dot([1, 1], w_hidden))
actual_output = np.dot(actual_hidden, w_output)
print('[1, 1]', actual_output)